Publications

We developed ARUA, a prototype that positions users as 'directors' of their own AI commentary, tailoring its social presence, emotional tone, and cognitive load.

Studying how sycophantic AI behavior impacts the quality of value-laden deliberation and moral reasoning.

Investigating how visual information creates cross-modal interference in speech evaluation by multimodal LLMs, revealing cultural asymmetries.

Uncovering gender matching biases in multimodal AI teammate configurations.

A benchmark for evaluating how well multimodal models ground acoustic cues versus text in medical triage.

Examining how persona designs for AI shopping agents influence users' cognitive and social engagement.

Revealing authorship biases in how users evaluate generated content in IR systems.

Using a collaborative cooking game (Co-Overcooked), we show that partner modeling capacity constrains viable human-AI team configurations in nonlinear ways, revealing expectation conflicts when multiple humans hold divergent models of a shared AI partner.

Designed a fine-tuned LLM-based AI agent for sports co-viewing that significantly enhances users' emotional engagement.

A novel multi-agent system for film appreciation where users converse with AI personas of directors, actors, and audiences.

Revealing how contextual cues create halo effects in multimodal LLM-based hiring systems.

A benchmark for evaluating LLM tool-use capabilities in complex multi-party dialogue scenarios.

Developing strategies to counteract selective exposure in conversational argument search.

Investigating how user participation in chatbot customization affects tolerance of chatbot failures.

Investigating how self-reference effects influence review behavior and evaluation quality.

Exploring how LLMs make decisions under pressure, revealing systematic biases in high-stakes scenarios.

Combining node and edge representations for inductive learning on large heterophilic graphs.

Addressing missing node features in GNNs by leveraging multi-hop neighborhood information.